Kozminski Techblog

Reading Time: 2 minutes

While this post blog is being written electrons are responsible for computing and photons are used for carrying data worldwide. Before optical fiber’s first implementation electrons had been used for both computing and transmitting tasks. It is viable for photons to take also computing tasks over electrons, which can lead to the development of optical computers. What are the advantages and limitations of optical computing and how does it differ from a conventional one, you will find out the answer after reading this post.

Difference between optical and electronic computing: Optical computers use photons to transmit data and perform operations, while traditional electronic computers use electrons to transmit data and perform operations.

Difference between photons and electrons:
Photons are electrically neutral, which enables crossing each other without interacting. On the contrary, protons don’t have this feature, which creates the main difference between them. It means that glass fibers can handle many simultaneous signals in a way that copper wires cannot. That is why optical computing can potentially operate at much faster speeds than electronic because photons can transmit data much more quickly than electrons, moreover, it solves the issue of parallelism. It enables performing more than one calculation simultaneously, which is impossible in the case of electronic computing. This could make optical computers much more efficient and faster at performing certain types of calculations. What types of calculations? Optical computing is efficient in resolving calculations of linear algebra and profoundly worse at different types of calculations. But taking into account that linear algebra is fundamental to machine learning so generally for AI, we can assume that the technology can bring many advantages in the future.

Advantages of Optical technology:

• Using photons reduces power consumption compared with the usage of protons. It is associated with the electrical resistance of protons, which generates heat and, wastes energy. The passage of photons through transparent media is resistance-free.
• Unlike standard computers, which only can make one calculation simultaneously, optical computers could do lots of calculations at the same time.

Potential Usecases:

• It is possible to implement it in devices that cannot be equipped with electronic chips because of the low capacity of the battery, owing to less energy consumption optical chips could eliminate the need for communication between a device and server, which is responsible for making complex computing and sending a response back to the device, which is time-consuming and insecure.
• Optical computing is able to handle big data sets, which would enable to work with high-resolution pictures in facial recognition and object detection processes, which makes them faster and much more accurate.

Disadvantages of optical computing:

• Complexity: Building optical computers can be technically challenging because it requires the development of specialized components and materials.
• Limited applications: Optical computers may not be suitable for all types of applications, because they may not be able to perform certain kinds of calculations as efficiently as electronic computers.

Summary: All things considered, optical technology is in the early stage of development. With new improvements and ideas to enhance its diversity of use and production of optical computers, it will gradually be taking over conventional technology. Assuming, that the demand for computing power enormously has increased and will increase, and AI applications have required and will require better efficiency, we can be almost sure that in the near future optical technology so also optical computers or hybrid ones will be desired by markets.

Sources:
https://www.economist.com/science-and-technology/2022/12/20/artificial-intelligence-and-the-rise-of-optical-computing

Reading Time: 2 minutes

Python is one of the most popular programming languages for data scientists, thanks to its powerful libraries and extensive support for machine learning and other data analysis tasks. There is a roadmap that I’ve gone through and recommend to you in order to get a basic understanding of Data Science from a programmatic perspective.

Step 0: Learn basic Python concepts:
Familiarizing yourself with the basics of programming. This includes learning fundamental concepts such as variables, if statements, loops, functions, and classes, also understanding basic data structures in Python, such as lists, dictionaries, and tuples.

Step 1: Perform basic operations on datasets

• Numpy
• Pandas

Numpy and pandas are two of the most popular libraries for working with data in Python. Numpy is a library designed specifically for numerical computing, providing powerful tools for performing complex mathematical calculations. Pandas, on the other hand, provides a more general-purpose framework for data analysis tasks, offering tools for exploring and visualizing data as well as for performing more advanced data wrangling and analysis.

Step 2: Visualize your data

• Matplotlib
• Plotly

Matplotlib provides tools for creating and visualizing data plots, while Plotly is a more advanced library that offers powerful tools for producing interactive data plots. Both of these libraries are widely used by data scientists working with Python, offering easy-to-use APIs for creating rich data visualizations.

Step 3 Begin your journey with basic algorithms of machine learning

• Scikit-learn

SciKit-Learn is a machine learning library that provides support for common algorithms like regression, classification, and clustering, making it easy to train models or perform data analysis tasks with just a few lines of code. SciKit-Learn also has tools for preprocessing data, visualizing the results, and testing models.

Step 4: Dive into the world of deep learning

• TensorFlow
• PyTorch

TensorFlow and Pytorch are two of the most popular libraries for performing machine-learning tasks in Python. TensorFlow was developed by Google and is a powerful, general-purpose library that supports a wide variety of machine-learning algorithms. It offers tools for training neural networks and other deep learning models, as well as support for mathematical operations on matrices and arrays. Pytorch is a newer library that was developed by Facebook, and it offers similar functionality to TensorFlow but with a more flexible, dynamic architecture that makes it easier for developers to customize and tailor their models for specific applications.

Step 5: Deploy your model

• Flask
• Django

Flask and Django provide tools for building web applications. Flask is a lightweight library, designed to be quick and easy to use, while Django is a more comprehensive framework, with many built-in features and utilities that make it easier to manage complex web applications. Both libraries provide powerful templating capabilities and other essential tools for developing web applications in Python.

Summary:
There are a lot more libraries to learn as a data scientist to face real-world tasks and meet expectations. But these libraries are powerful and can be qualified as the main ones after learning that you are a valuable candidate in the job market. Personally, it took me half a year to feel comfortable with using libraries like Numpy, Pandas, Matplotlib, Plotly, and Django. Other libraries, especially, these about machine learning require more time, practice, and basically more theoretical knowledge to become effective and confident with using them. These libraries are essential to become good at Data Science, other libraries, which you will learn along the way, are just an addition to the core that incorporates libraries included in the article that you’ve just read.
Thank you for your time!

Sources:
https://builtin.com/data-science/python-libraries-data-sciencehttps://pandas.pydata.org/docs/index.html https://numpy.org/doc/stable/
https://matplotlib.org/ https://plotly.com/ https://www.djangoproject.com/start/overview/

Reading Time: 2 minutes

While I was reading an article in Financial Times “Has relying in tech made us more stupid” written by Dave Lee, I dived into inspiration to make research on detrimental effects of GPT-3-like AI. The results of it can be read below:

AI is great at providing us with quick answers and accelerating a creative process, but there is the first jeopardy, namely, it can limit our thinking in many ways. AI applications frequently give us familiar results, even when we ask slightly different questions. This means that we are less likely to find out of new ideas or explore alternative solutions because AI has already provided us with a standard answer pushing us in a limited number of directions.

Moreover, AI can be very limited in its scope, meaning that it may not always provide the most accurate information or biased one. This can reduce our ability to think deeply about a problem or question and come up with creative solutions ourselves and also it can prevent us from thinking more critically about the topics on which we’re doing research and exploring other perspectives.

That brings about another plausible threat, namely, the comfort of using GPT-3 AI can lead people to rely on it too much, forgetting how to think critically and solve problems independently. AI can also make us less patient, as we may become accustomed to relying on AI for quick answers and having our problems solved in an instant.

Last but not least, AI is poor at thinking abstractly as humans can, so it is unable to come up with creative solutions or jokes that take a different approach.

Key Takeaways:

• Exceeded reliance on AI may lead to weakening critical-thinking skills
• Disability of AI to think abstractly may lead to limiting human creativity
• AI may be biased and limited toward certain outcomes, leading to rooting stereotypes and limiting diversity.

This technology is rapidly evolving, so it’s important to stay informed on the latest developments and learn how they can help in our daily life. Personally, GPT-3-like AI will take the world by storm, helping people with immediate access to endless data and accelerating the creative process, nevertheless, over-using and over-reliance on the AI can lead to disappearing vital skills of humanity, for example, critical thinking. While AI can be an incredibly helpful tool, it’s important to remember that it should not replace our own intelligence. As Paracelsus said  –

“Everything is poison and nothing is poison because only the dose makes the poison”

References:

https://www.ft.com/content/f9e3e5c8-e634-4c63-aaf3-6d16987278e7 https://bernardmarr.com/what-are-the-negative-impacts-of-artificial-intelligence-ai/
https://www.univ.ai/post/the-limitations-of-gpt-3-and-its-impact-on-society

Reading Time: 2 minutes

Having the third decade of the twenty-first century, the scientific world has constructed at least a dozen worth-knowing humanoid robots. Sophia, Ameca, Amelia, Pepper, Beomni, or Atlas these ones are names of the robots that we have available on the market today, among others. In this article, you can read my synthesis of these particular projects and you can find references to videos presenting their performance. These six projects, as far as I know, are good benchmarks for the others currently available on a market.

(Category: Human-oriented)
Sophia by Hanson Robotics
&
Ameca by Engineered Arts

Undoubtedly, these are the best robots resembling human appearance. Sophia and Ameca are characterized by sophisticated interactivity. Besides looking like a human they behave like a human. Advanced facial expressions, understanding a natural language, and responding to it, and body movement are basic features that these robots can perform. The best way to get a better understanding of their skills is to see them in action, and you can do this through the following links:
Ameca: https://www.youtube.com/watch?v=EWACmFLvpHE
Sophia: https://www.youtube.com/watch?v=JRHdnkUjcZg&t=2s

(Category: Movement-oriented)
Atlas Boston Dynamics

The most sophisticated in terms of reflecting a human movement. Jumping, squatting, pushing, dancing, or even flipping are not activities that Atlas cannot perform. It has been developed to revolutionize the capabilities of robotic movement. It can perform parkour you can see this through the following link:
Atlas: https://www.youtube.com/watch?v=tF4DML7FIWk

(Category: Utility-oriented)
Beomni 1.0 by Beyond Imagination
&
Pepper by Humanizing Robotics

These two robots share most of the features that characterized Ameca and Sophia robots apart from the following one: Reflection of human appearance. These ones are available to buy, which means that you can take advantage of them even tomorrow (Beo. They have been used in a few industries so far. This feature gives their producers a competitive advantage over the owners of robots that I wrote about before. So let’s see their performance through the links below to figure out their possible application in your life:
Beomni: https://www.youtube.com/watch?v=fjC4PBj6SKU
Pepper: https://www.youtube.com/watch?v=Ti4NiaQj8q0

(Category: Virtual-oriented)
Amelia by Amelia Ai

Amelia differs from previous ones because is the only one that is fully digital one. Amelia is the most advanced conversational AI, that is known as a successor of chatbots. It is a highly sophisticated virtual assistance that is used for customer services. The neural networks implemented in Amelia are comparable to those ones that are implemented in Ameca and Sophia. But this project is worth highlighting because it shows other filed of AI utilities. The functionalities of Amelia you can check via the link below. Amelia: https://www.youtube.com/watch?v=B9u3yiJGmSI

Summary of the article
Hopefully, you’ve brought something new for you with this short article, my main purpose to write it has been to collect links and make it easier to you to catch up with news on humanoid robots. Personally, I reckon that human-like machines are our future, which will bring many opportunities for humanity, for example, providing accompany for solitude people helping them with simple daily chores, holding conversations to overcome loneliness, or providing customer services for businesses. We need them, and it depends on us how we will use them. If creators take responsibility for developing them in a direction of helping out rather than reducing the freedom and privacy of people, it will take the world by storm in the following decade.
Thank you for your time

References:
https://www.hansonrobotics.com/sophia/
https://www.engineeredarts.co.uk/robot/ameca/
https://www.bostondynamics.com/atlas
https://humanizing.com/en/
https://www.beomni.ai/
https://amelia.ai/

Reading Time: 2 minutes

At the beginning of Ukraine’s defense against the Russian invasion, the technology that has been provided by SpaceX called StarLink enables ubiquitous access to the satellite internet connection, which gives a tactical advantage of surveillance of the battlefield for Ukraine’s troops over Russian ones who struggle with poor internet connection. But what would happen if SpaceX stopped providing this game-changer technology?

According to the latest news, Elon Musk the CEO of SpaceX has been accused of holding a conversation with Mr. Putin. After he tweeted a question if his followers approved of the proposal to end the war in Ukraine under Rusian’s conditions which contain, among others, the approval of Rusian’s annexation of Crimea.

To predict what would happen if Ukraine’s army became StarLink deprived, let’s refer to issues that StarLink has been a response for, and its implementations as well:

Vulnerability of internet infrastructure:

• Since the first days of the Russian invasion, Russian forces have attacked crucial infrastructure responsible for connectivity:
  • Attack on television towers in Kyiv on Feb. 24th.
  • Attack on television towers in Kharkiv on Feb. 24th.
  • Bombarding offices of Kyivstar in Mariupol on March 21st. (Ukraine’s largest internet provider)
  • Successful attempt to hack Viasat on Feb. 24th.
    (Viasat is an American provider of a network that was used by the Ukrainian troops to exchange information between coordinating units and with the army located on the battlefield)

Implementation of StarLink:

• Tracking targets for drone strikes.
• Daily communication between Unkrain’s units.
• Communication bridge between Ukraine’s soldiers and their families.
• Keeping Ukraine’s population on the ball, providing reliable information from the Western media.
• Enabling to share of Ukraine’s President’s speeches on the battlefield, in order to rise the spiritual morals of Ukraine’s troops.

If we take these implementations and the vulnerability of Ukraine’s infrastructure and turn them upside down, we will see a few underbellies that Ukraine would be exposed to in case of a lack of StarLink technology.

To sum up this post. I will use one of the Ukrainian soldier’s quotes “Starlink is our oxygen, you can’t just turn it off”

Thank you for your time.
Feel free to add comments below.

Sources:

https://www.politico.com/news/2022/06/09/elon-musk-spacex-starlink-ukraine-00038039

https://www.economist.com/graphic-detail/2022/04/29/satellite-internet-is-a-hot-new-commodity-in-ukraine

https://www.economist.com/briefing/2022/10/06/elon-musks-foray-into-geopolitics-has-ukraine-worried